You are on page 1of 10

Received: 16 September 2017 | Revised: 20 June 2018 | Accepted: 30 June 2018

DOI: 10.1111/jbi.13422

RESEARCH PAPER

Islands in the trees: A biogeographic exploration of epiphyte‐


dwelling spiders

Francisco Emmanuel Méndez-Castro1,2 | Maaike Y. Bader1 | Glenda Mendieta-Leiva1 |


Dinesh Rao2

1
Ecological Plant Geography, Faculty of
Geography, University of Marburg, Abstract
Marburg, Germany Aim: Epiphytic plants are isolated from each other by nonhabitat canopy elements
2
Instituto de Biotecnología y Ecología
and are thus expected to act as islands, the biodiversity of their inhabitants (e.g.,
Aplicada (INBIOTECA), Universidad
Veracruzana, Xalapa, México spiders) conforming to island biogeographic predictions of species‐richness patterns.
Although it has been shown that arthropod diversity decreases with decreasing epi-
Correspondence
Francisco Emmanuel Méndez-Castro, phyte size, the effects of isolation have not yet been addressed. We studied the
Deutschhausstraße 10, 35032 Marburg,
joint effect of isolation, spatial position, and size of epiphytic plants (canopy islands)
Germany.
Email: femendez@icloud.com on species richness, relative rareness, and similarity of spider communities.
Location: A shade‐coffee plantation in central Veracruz, Mexico.
Funding information
Universidad Veracruzana, Grant/Award Taxon: Spiders (Araneae), vascular epiphytes (Bromeliaceae, Piperaceae, Orchi-
Number: PROMEP/103.5/10/7748;
daceae, Araceae, Pteridophyta).
CONACYT, Grant/Award Number: 307402/
251146 Methods: We collected all canopy islands occurring on three trees and recorded
their three‐dimensional spatial position. In the laboratory, we disassembled the
Editor: Triantis Kostas
plants and collected all spiders present. We analysed the effects of island size, isola-
tion, and spatial position on the species richness and the relative rareness of spider
communities using generalized linear models (GLM), on the distribution of different
spider guilds using a CCA, and on community composition using a permutational
multivariate analysis of variance (Permanova). Isolation and spatial position were
addressed using five distance measures representing isolation from different poten-
tial species sources. We tested how the similarity in spider community composition
changed with the distance between islands using Mantel correlograms.
Results: As predicted by island biogeography theory, spider species richness
increased with canopy‐island size and decreased with isolation. In comparison to
island size, the effect of isolation was weak, though significant. Relative rareness
was hardly affected by island size but more by isolation. Compositional similarity
was affected by island size and decreased with increasing spatial distance up to ca.
4 m. Guilds separated along the main CCA axes, this ordination being driven by epi-
phyte size and position.
Main conclusions: Epiphytic plants behaved like canopy islands in that their size
and isolation influenced the diversity and composition of spider communities. How-
ever, the effect of isolation was only a fraction of that of island size, perhaps
because spatial relationships are taxon‐specific. This may be due to differences in
hunting behaviour and dispersal capacities, for example, between guilds of hunting

Journal of Biogeography. 2018;1–10. wileyonlinelibrary.com/journal/jbi © 2018 John Wiley & Sons Ltd | 1
2 | MÉNDEZ‐CASTRO ET AL.

and web‐building spiders, which is a dimension deserving more attention. For a bet-
ter understanding of biogeographic principles driving the diversity of canopy island
inhabitants, further research on this topic should include position and isolation, at
scales matching the mobility of different functional groups, as part of their explana-
tory variables.

KEYWORDS
epiphytes, habitat islands, Island biogeography, spatial patterns, species richness, spiders

1 | INTRODUCTION Nieto, & Romero, 2011; Méndez‐Castro & Rao, 2014; Pinzon,
Spence, & Langor, 2013).
According to the equilibrium theory of island biogeography (ETIB), In this study, we analysed the combined effect of epiphyte size,
species numbers on islands are primarily determined by the interac- spatial position and isolation on the species richness, the relative pro-
tion of island size and isolation (Cox, Moore, & Ladle, 2016; portion of rare species, and the community similarity of epiphyte‐inha-
MacArthur & Wilson, 1967; Whittaker & Fernández‐Palacios, 2008). biting spiders. Spiders and epiphytes have a mutualistic relationship in
Larger and less isolated islands are expected to harbour more spe- which the plants provide spiders with a suitable habitat for foraging,
cies than smaller and more isolated ones. Islands do not need to be reproduction, egg laying, and finding shelter (Gonçalves‐Souza,
surrounded by water; any habitat patch embedded in a nonhabitat Brescovit, Rossa‐Feres, & Romero, 2010; Scheffers, Edwards, Dies-
matrix can be considered an island and species numbers in such mos, Williams, & Evans, 2014), and spiders reduce herbivore popula-
habitat islands should conform to the size–isolation pattern predicted tions through predation and can make nutrients available to the plants
by the ETIB (Adams, Schnitzer, & Yanoviak, 2016; Öckinger, Franzén, (e.g., via excretion) (Gonçalves, Mercier, Mazzafera, & Romero, 2011;
Rundlöf, & Smith, 2009; Prugh, Hodges, Sinclair, & Brashares, 2008). Romero, Mazzafera, Vasconcellos‐Neto, & Trivelin, 2006). Spiders are
Epiphytic plants in the canopy of tropical forests can be considered highly sensitive to subtle changes in habitat structure (Gonçalves‐
as islands since they are discrete habitats separated from each other Souza et al., 2011) and microclimatic conditions and they show active
by a dissimilar matrix. These canopy plants form habitat islands for habitat selection (Omena & Romero, 2008; Rao, 2017). Because of this
many arthropods and other organisms (Richardson, 1999) and are sensitivity, habitat diversity is expected to positively affect spider
arranged in “archipelagos” in a three‐dimensional and discontinuous species numbers even at small spatial scales (Aikens & Buddle, 2012;
habitat space (Bader, van Dunné, & Stuiver, 2000). Barton et al., 2017; De Mas, Chust, Pretus, & Ribera, 2009). On top of
In accordance with the ETIB, larger epiphytes have been found being selective and sensitive towards the physical structure of their
to host a greater number of arthropod species such as ants, beetles, habitat, spiders display a variety of foraging strategies and can be
springtails, and spiders (Méndez‐Castro & Rao, 2014; Rodgers & classified into functional guilds (Cardoso, Pekár, Jocqué, & Coddington,
Kitching, 2011; Stuntz, Ziegler, Simon, & Zotz, 2002; Wardle et al., 2011), which differ in their mobility (e.g., web builders vs. active
2003). However, the effect of isolation on faunal diversity is difficult hunters). These differences are expected to work as behavioural filters
to demonstrate (Barton, Evans, Foster, Cunningham, & Manning, modulating the effects of island size and isolation on spider diversity
2017; Jiménez‐Valverde, Baselga, Melic, & Txasko, 2009; Wu, Si, at the community level (Cobbold & MacMahon, 2012).
Didham, Ge, & Ding, 2017). So far, isolation effects (interisland dis- Although spiders can move between epiphytes along branches
tance) on canopy fauna have been studied only once, for canopy soil or, depending on the species, through the air (Mestre, Bucher, &
communities in real and artificial soil accumulations in New Zealand, Entling, 2014; Wolff, Schneider, & Gorb, 2014), the space between
where no significant effects were detected (Wardle et al., 2003). the epiphytes contains only a fraction of the spider occurrences
For most islands, isolation is defined as the linear distance(s) of a observed in epiphyte‐rich trees (Angelini & Silliman, 2014; Cruz‐
given island to one or several species sources (mainland or neighbour Angón, Baena, & Russell, 2009) and is therefore considered as a
islands). Unlike distances between real islands and terrestrial habitat nonhabitat matrix in this study.
fragments, distances between canopy islands are defined in a three‐ We aimed to test to what extent the predictions of the ETIB with
dimensional space (Bader et al., 2000). It is unclear whether isolation respect to size and isolation effects on biodiversity are valid for spider
within this space plays a role in determining species distributions. communities at the spatial scale of canopy ecosystems. We hypothe-
One reason for this uncertainty is the multitude of potential species sized that: (a) epiphyte size will have a positive effect on spider species
sources, such as neighbouring epiphytes, the understorey or the sur- numbers, (b) isolation will have a negative effect on spider species
rounding trees. The effect of epiphyte spatial distribution has been numbers and community similarity, (c) more isolated epiphytes will
addressed previously, but only considering vertical stratification, contain a higher proportion of species that are rare in the overall sam-
which is generally found to be important for the composition of pling, and (d) spiders will show specific responses towards island fea-
canopy fauna (Aikens & Buddle, 2012; Gonçalves‐Souza, Almeida‐ tures depending on their hunting strategies (guilds).
MÉNDEZ‐CASTRO ET AL. | 3

2 | MATERIALS AND METHODS 2.2 | Analysis

2.1 | Data collection We evaluated what proportion of the locally available spider species
pool was represented in the three trees by means of the interpola-
The study area is located in southeast Xalapa, Veracruz, Mexico. The tion and extrapolation analysis described by Chao et al. (2014),
site is a 40‐year‐old shade‐coffee plantation called “La Orduña” which assesses sample completeness based on sample coverage. We
(19°28′02″N, 96°55′45″W), a well‐known site for the study of spe- used the function iNEXT from the R package ‘iNEXT’ (Hsieh, Ma, &
cies diversity (Cruz‐Angón et al., 2009; Manson, Hernández‐Ortiz, Chao, 2016).
Gallina, & Mehltreter, 2008). We selected three trees of the domi- The full epiphyte and spider data sets included four dependent
nant species Inga jinicuil G. Don (Mimosaceae), which harbours the and 10 independent variables, out of which a subset was used for
greatest diversity of epiphytes (more than 40 species; Cruz‐Angón et analysis (data summarized in Tables 1 and 2, variable descriptions
al., 2009). Epiphytes and spiders were collected as follows: for the subset below and for the rest in Appendix S1). We tested
the collinearity of the variables using pairwise nonparametric
1. We climbed the trees using the single rope technique (Perry, 1978). Spearman correlations (corr.test function of the R package ‘psyc’;
2. We recorded three-dimensional epiphyte positions: “x” and “y” Revelle, 2017). When variables strongly correlated (r > |0.5|), we
coordinates were measured along measuring tapes, with the retained only those we considered most informative and discarded
trunk of the tree as the origin (Figure S1 in Appendix S1), and those that were redundant (Table S2 in Appendix S1). From the
the position on the “z” axis (corresponding to the distance from full set of four dependent variables (abundance, species richness,
the ground, as the trees were located on flat terrain) was mea- functional diversity, and relative rareness), we retained only spe-
sured by dropping down a measuring tape. cies richness, as the most basic diversity measure, and relative
3. We carefully collected canopy islands by removing the roots of rareness, which was least correlated (r ≤ |0.30|). In the case of
the epiphytes from the tree, after which each plant was wrapped independent variables, all island size and diversity variables (epi-
in a plastic bag to avoid the loss of mobile spiders (Stuntz et al., phyte abundance, epiphyte dry weight (log island size), epiphyte
2002; Yanoviak, Nadkarni, & Gering, 2003). groups, and habitat diversity index, Table S2 in Appendix S1)
4. We disassembled canopy islands in the laboratory and captured showed strong correlations with r > |0.6|, so that we retained only
all spiders. We collected spiders by hand and preserved them in island size. The spatial variables (branch diameter, distance from
80% ethanol for further examination. We sorted collected spiders the ground, horizontal distance, average isolation, nearest neigh-
into morphotypes according to their morphological traits and, bour distance, and nearest neighbour index, Table S2 in
when possible, we identified adults to species following taxo- Appendix S1) correlated only weakly, with r < |0.30|, except
nomic keys available in the literature (Jocqué & Dippenaar- branch diameter with distance from ground (r = −0.51) and hori-
Schoeman, 2006; Ubick, Paquin, Cushing, & Roth, 2005; World zontal distance (r = −0.69). Therefore, all spatial variables except
Spider Catalog, 2018). Spider vouchers are located at the Labora- branch diameter were maintained.
torio de Aracnología de la Facultad de Ciencias, Universidad
Nacional Autónoma de México (UNAM).
5. We dried the epiphytes for two days at 70°C and weighed them 2.2.1 | Dependent variables used
(Affeld, Sullivan, Worner, & Didham, 2008). Spider species richness corresponds to the number of species per
canopy island. The relative proportion of rare spiders per island was
Before collecting epiphytes (or “canopy islands”), we conducted a determined using the “Index of Relative Rarity” (hereafter referred
pilot analysis (nearest‐neighbour clustering) on 17 Inga jinicuil trees to
identify those with the highest structural similarity. Variables consid-
T A B L E 1 Summary of tree, canopy island (epiphyte), and spider
ered were diameter at breast height, number of branches, epiphyte
community characteristics for three trees in a shade‐coffee
load, and canopy cover. Trees analysed were at least 50 m apart from plantation in Mexico. Canopy islands could occur either as epiphyte
each other to ensure independence. Based on our previous experience mats composed of several plants or as single plants
collecting epiphyte‐dwelling spiders, spiders can last in the collection Tree Tree 1 Tree2 Tree3 Total
bags up to five days without getting damaged or suffocating (Méndez‐
Stem DBH (cm) 52 51 59
Castro & Rao, 2014). Therefore, we collected a maximum of 50 epi-
Tree height (cm) 1480 1480 1260
phytes per day to have enough time to process the samples and to
Maximum tree‐canopy radius (cm) 432 623 780
avoid accumulation of unprocessed samples. Because of the high num-
Epiphyte individuals 363 314 514 1191
ber of epiphytes per tree (Table 1), it took us about seven weeks to
Canopy islands 120 83 96 299
collect ca. 95% of the epiphytes occurring on a single tree. The remain-
Total spider species richness 94 89 99 164
ing 5% of plants corresponded to epiphytes found on the thin
branches of the outer canopy, which were not collected for safety rea- Total number of rare species 64 56 58 127

sons. We collected epiphytes from October 2013 to March 2014. Total number of spiders 2631 2736 2113 7480
4 | MÉNDEZ‐CASTRO ET AL.

T A B L E 2 Summary of canopy island (epiphyte) and spider any of them were collected. We tested whether this might have
community characteristics for the canopy islands on three trees in a masked isolation effects (due to communities reaching new equilibria
shade‐coffee plantation in Mexico. To determine the epiphyte group in between sampling campaigns) using a timed average isolation (tai),
diversity, we defined six groups of epiphytes according to their
calculated by excluding those islands that were removed by previous
taxonomy: Bromeliaceae, Orchidaceae, Pteridophyta, Piperaceae,
Araceae, and miscellaneous (some epiphytic cacti and hemiepiphytic sampling from the calculation of average isolation. If there would be
Ficus). Habitat diversity is an index that combines island size and the a significant effect of sampling time, we would expect tai to have a
number of epiphyte groups higher explanatory power than ai.
Canopy island Min Mean Max We ran generalized linear models (GLM) to analyse the effect of

Number of epiphytes 1 5 40 canopy‐island size and spatial variables on spider species richness

Epiphyte group diversity (egd) 1 2 6


and relative rareness. The initial models included island size, distance
from the ground, horizontal distance, average isolation, nearest
Spider species richness 1 8 37
neighbour distance, and nearest neighbour index and all two‐way
Number of spiders 1 24 487
interactions. Tree identity was included as a covariable to account
Branch diameter (cm) 2 14 59
for intertree variation. Additionally, for tree1, we calculated these
Canopy‐island size (cis, in g) 1 170 4222
models with ai replaced by tai. To find the minimal set of indepen-
Habitat diversity (log [(cis +1) * egd]) 0.3 3.4 18.1
dent variables explaining the highest amount of variation, we used
Nearest neighbour distance (cm) 10 70 233 stepwise backwards model selection (drop1 function) to remove the
Average isolation (cm) 369 524 990 least significant variables. The percentage of explained variance (D2)
was calculated by dividing the residual deviance of each term (pro-
vided by the function anova) by the total deviance of the model
as “relative rareness”), which was calculated by assigning weights to (Kindt & Coe, 2005).
spider species based on their abundance in the total sample and To assess how the similarity (Bray–Curtis) between spider com-
then summing these weights per island. These summed weights are munities correlated with the Euclidean distance between canopy
then divided by the species richness of each canopy island so that islands, we computed Mantel correlograms for each tree. The dis-
the proportion depends on the number of rare species but not on tance classes were set at 100‐cm intervals and we used only dis-
species richness (Leroy, Petillon, Gallon, Canard, & Ysnel, 2012). tance classes that included comparisons between all canopy islands
(distances above 700 cm were omitted). We used Spearman correla-
tions and 999 permutations. Using an analysis of similarity (function
2.2.2 | Independent variables
anosim), we compared dissimilarities in spider species composition
Epiphyte biomass (dry weight) was used as the measure of island between islands among and within trees, and with a permutational
size. Euclidean distances between the epiphytes were calculated multivariate analysis of variance (Permanova, function adonis2), we
using the function vegdist of the R package ‘vegan’ (Oksanen et analysed which of the tree and canopy island features explained
al., 2017). The spatial multidimensionality of the canopy was these dissimilarities. Both anosim and permanova use dissimilarity as
addressed by using five distance variables (Figure 1): (a) distance the default distance measure. We explored the specific responses of
from the ground (dg) corresponds to the distance on the “z” axis spider guilds to island features using a canonical correspondence
(height) and represents isolation from the understorey; (b) horizon- analysis (CCA) based on the species per guild, plotting the guilds,
tal position (hod) represents isolation from the centre of the tree and independent variables. Guilds were assigned to spider species
(corresponding to the zone with the largest epiphyte biomass; according to their family (Table S2.2 in Appendix S2) following the
Freiberg & Freiberg, 2000; Díaz, Sieving, Peña‐Foxon, Larraín, & guild classification proposed by Cardoso et al., (2011). Anosim,
Armesto, 2010; Nakanishi, Sungpalee, Sringernyuang, & Kanzaki, Permanova, correlograms, and CCA were all calculated using the
2016) (c) nearest neighbour distance (nnd) is the Euclidean dis- R package ‘vegan.’
tance from a given canopy island to its closest neighbour and rep- All analyses were performed in the statistical software R version
resents isolation from the nearest potential species source; (d) 3.3.2 (R Development Core Team, 2015).
nearest neighbour index (nni) represents edge to the edge (rather
than centre to centre) Euclidean distance between canopy islands
and was calculated as nni ¼ pnnd
ffiffiffiffiffiffiffiffi, where nns = nearest‐neighbour
nns 3 | RESULTS
size and nnd = nearest‐neighbour distance (adaptation of an idea
proposed by Gustafson & Parker, 1994); and (e) average isolation We collected 1,191 epiphytic plants clustered in 299 canopy islands
(ai) is the average distance of a given island to all other islands in on three trees (Table 1). We extracted 7,480 spiders and sorted
the same tree and summarizes the spatial position of canopy them into 164 morphospecies representing 48 genera and 26 fami-
islands in the tree. lies (Table S2.1 in Appendix S2). Trees harboured about 98% of the
Isolation variables were calculated considering the position of species likely to be present in the area while each tree harboured
the canopy islands in their initial configuration on the trees, before over 90% (Figure S3 in Appendix S1).
MÉNDEZ‐CASTRO ET AL. | 5

Axis origin

1200 Horizontal distance (hod)


3

1000 Nearest neighbor distance (nnd)

5
800

Distance to ground (dg)


Z

600
6
4

400 35 30 25 20 15 10 5

2 Spider species richness


7
200
Branch diameter
1 Spider species richness
600
400
400 200
200 0
0 -200
-200 -400 X
Y
-400 -600
-600

F I G U R E 1 Three‐dimensional scatter plot of the spatial position of canopy islands (composed of epiphytes) in a tree (tree 3) in a Mexican
coffee plantation, showing the meaning of different distance and isolation metrics. Canopy islands are represented by bubbles, bubble size
(legend 1) represents island size, and bubble colour (legend 2) refers to spider species richness. Horizontal distances to the trunk (arrow 3)
were calculated taking the axis origin as a reference point. Average isolation (inset 4) is the mean Euclidean distance from a given canopy
island to all other islands occurring on the same tree. In most cases, the nearest neighbour (arrows 5) of given canopy islands was located on
the same branch, but it could also be on a neighbouring branch. Distance from the ground (arrow 6) and branch diameter (bars 7) were
measured at the same point (the middle of the epiphyte mats). The scale of the axes is in cm; canopy island positions were recorded with this
precision

The spider families with highest abundances were the Dictynidae for relative rareness (explaining 6.2% of the variation), while for spe-
(n = 1042), Pholcidae (n = 899), and Anyphaenidae (n = 897), while cies richness, it explained <1%. The third significant explanatory vari-
the most species‐rich families were Salticidae (23 spp.), Theridiidae able for relative rareness was average isolation, while for species
(14 spp.), and Anyphaenidae (12 spp.). Around 75% of the species richness, this variable was only significant in interaction with hori-
were classified as rare in the total sample. zontal distance (Figure S4B in Appendix S1), explaining only 1.7% of
Generalized Linear Models (Table 3) showed that although island the variation. Distance from the ground and horizontal distance (Fig-
size, distance from the ground, and average isolation were significant ure S4A in Appendix S1) also significantly interacted in the species
drivers of spider species richness and relative rareness, their contri- richness model, but again with little explanatory power (0.6% of vari-
bution to the amount of explained variation differed according to ation explained). Models using timed average isolation (Table S6 in
the dependent variable. Overall, the models could explain 77% of Appendix S1) retained the same set of explanatory variables as mod-
the variation in species richness and just 13% of that in relative rare- els using average isolation (Table 3). While timed isolation produced
ness. Island size explained about 68% of the variation in species rich- no improvement on the explanatory power of species richness
ness (Table 3, M1) but only 3.1% of the variation in relative rareness model, it doubled the amount of explained deviance for the model
(Table 3, M2). Only in the model for relative rareness, island size of relative rareness. However, timed isolation itself did not have a
interacted with average isolation (Figure S4C in Appendix S1). Dis- significant effect but only modified the effect of distance from the
tance from the ground was the most important explanatory variable ground through an interaction. Therefore, we concluded that using
6 | MÉNDEZ‐CASTRO ET AL.

T A B L E 3 Generalized linear models of spider diversity on epiphytic plants in a Mexican shade‐coffee plantation. The two models differ in
their dependent variables (M1: species richness, M2: relative rareness). Shown are model estimates, p‐values, and the percentage of explained
variation (%D2) for each variable. We used the quasipoisson family for both models. Even though all combinations of island and spatial
variables and their two‐way interactions were tested, the minimal models presented here contain only the significant variables selected using
stepwise model selection
M1 M2

Species richness Relative rareness

z value Pr(z) %D 2
z value Pr(z) %D2
(Intercept) 54.7 *** – (Intercept) −32.8 *** –
Log island size (lis) 29.2 *** 68.5 Log island size (lis) 1.8 . 1.0
Distance from the ground (dg) −3.8 *** 0.8 Distance from the ground (dg) −3.9 *** 6.2
Horizontal distance (hod) −1.1 – 2.1 Average isolation (ai) 2.5 * 2.9
Average isolation (ai) 1 – 0.0 lis : ai −3.3 ** 3.1
Tree2 −7.1 *** 3.4 Total %D2 13.2
Tree3 −6.6 ***
dg : hod −3.9 *** 0.6
hod : ai −4.9 *** 1.7
Total %D2 77.1

Interactions are denoted by “:” between pairs of variables. Significance codes: <0.001 “***” 0.001 “**” 0.01 “*” 0.05 “–” >0.05.

timed isolation, that is, taking into account the changes in spatial 0.1
configuration caused by the sampling, did not help to understand
the ecological patterns in spider distributions. 0.05
Mantel correlation

The Mantel correlograms (Figure 2) showed that increasing dis-


0
tances between canopy islands decreased species similarity
(Bray Curtis) of spider communities (negative correlation) up to
-0.05
about 4 m. At larger distances, correlations turn positive and increas-
Tree 1
ing distances increase similarity. Two of the three trees conformed -0.1
Tree 2
to this pattern, but one showed a different result, with just one sig-
nificant negative correlation for distances around 1 m (Figure 2, -0.15
Tree 3
50 100 200 300 400 500 600 700
tree2). The analysis of similarity (anosim) revealed that within trees,
the mean dissimilarity (Bray–Curtis) between canopy islands was Distance class index
around 18%, which was significantly lower than the mean among
F I G U R E 2 Mantel correlograms based on the Bray–Curtis
trees 25% (Figure S5 in Appendix S1). According to the permuta- similarity of spider diversity and the Euclidean distances (in cm)
tional multivariate analysis of variance (Permanova), the dissimilarity between canopy islands (epiphytes) in three trees in a Mexican shade‐
in spider species composition between canopy islands was explained coffee plantation. Grey symbols represent nonsignificant correlations
by island size and tree identity (Figure S5 in Appendix S1). (p > 0.05)
Canonical correspondence analysis (Figure 3) organized guilds
into four groups suggesting specific responses towards island fea-
tures. Together, island size, distance from the ground, and horizontal 4 | DISCUSSION
distance explained 26% of the ordination of the guilds, with island
size as the dominant feature (variation explained, R2 = 0.22). The Trees can be islands for small cursorial animals, with crown isola-
CCA suggests that sheet and orb web weavers were the only guilds tion limiting free exchange between trees (Southwood & Kennedy,
sensitive to island size (y‐axis), orb web weavers having their highest 1983; Adams et al., 2016). In this system, epiphytes behave like
diversity in smaller islands than sheet web weavers. Space and sheet islands within islands (Richardson, 1999). Our data confirm that
web weavers and ambush and other hunters were most sensitive to spider communities on epiphytes conform to the expected positive
island position (x‐axis), space web weavers having their highest relationship between island size and species numbers and that iso-
diversity in islands closer to the ground and the main trunk than the lation is also a weak but significant driver of spider diversity and
other three groups. Ground hunters were the guild closest to the similarity at the community level. The importance of canopy‐island
centre axis of the ordination plot suggesting a relatively low sensitiv- size for arthropod species numbers and abundance has been
ity for island size or isolation. reported previously (Borges & Hortal, 2009; Méndez‐Castro &
MÉNDEZ‐CASTRO ET AL. | 7

-1.0 0.0 1.0

1.0
Log - island size

Sheet web weavers

0.5
Distance to ground

Space web weavers Horizontal distance


Ambush hunters

CCA2

0.0
0.0
Other hunters
Ground hunters

F I G U R E 3 Canonical correspondence
analysis showing the ordination of spider
guilds in canopy islands (epiphytes) in a
-0.5

Mexican shade‐coffee plantation according


to the effect of the size and position of Island features R2
the canopy islands. The effect of island Island size 0.22
position is summarized by the interaction

-1.0
Spaal posion 0.04
Orb web weavers
between distance from the ground and Residual 0.74
horizontal distance. R2 means proportion
-0.5 0.0 0.5 1.0
of variation explained
CCA1

Rao, 2014; Stuntz et al., 2002). However, in contrast to our Even though isolation had only a modest effect on spider species
results, the single other study addressing spatial distances among numbers, distances between epiphytes clearly affected the species
epiphytes did not show significant effects of interisland distances composition. As expected by the ETIB, nearby islands were more
or isolation from the ground on epiphytic soil communities (War- similar than islands at a larger distance from each other, up to a dis-
dle et al., 2003). tance of about 4 m. Apart from providing shorter dispersal distances,
Within a tree canopy, microclimate, substrate characteristics, short distances between islands also imply a similar environment,
and distance from the understorey vegetation, as a potential spe- which may also partly explain species similarity. This is illustrated by
cies source, all differ with height above the ground. Consequently, the increase in similarity at distances above 5 m, which is clearly not
invertebrate diversity is generally found to change along the verti- attributable to easy dispersal between islands but most likely to the
cal axis, but patterns are heterogeneous (Aikens & Buddle, 2012; similar microhabitat experienced at opposite positions in the outer
Gonçalves‐Souza et al., 2011; Pinzon et al., 2013; Rodgers & Kitch- crown (Scheffers et al., 2014; Wang et al., 2016; Woods et al.,
ing, 2011). For epiphyte‐dwelling spiders, the two available studies 2015).
related to vertical stratification report a decrease in species rich- The stronger effect of isolation (expressed as distance from
ness with height above the ground, tested up to 1.5 m (Gonçalves‐ the ground or distance between islands) on the relative rareness
Souza et al., 2011), and an increase with height above the ground, and on community similarity than on species numbers shows
tested up to 10 m (Méndez‐Castro & Rao, 2014). Our results, that spatial features affect the species composition even if the
including the full height of the trees, show that species richness overall number of species remains similar. One likely reason for
and the relative rareness per epiphyte island decreased with height the lack of overall isolation effects on diversity is indicated by
above the ground, although for species richness, this effect our ordination analysis, which points to spatial relationships
depended on the horizontal position. This could be interpreted as a being specific for functional groups. Spider communities are very
negative effect of vertical isolation. However, with only about 5% heterogeneous mixtures of species differing in dispersal mode,
contribution of the spatial variables to the total variance in species hunting behaviour, body size, and life‐cycle duration. According
richness, compared to 68% explained by canopy‐island size, the to our results, analysing this whole mixture as one could mask
role of spatial gradients within the canopy appears to be smaller responses of guilds or other functional or taxonomic subgroups
than we expected. This is particularly surprising since trees tend to to isolation and island size. At the small scale of epiphytes
have different epiphyte communities in different parts of the crown within a tree, island size seems to have a more pronounced
(Wang et al., 2016; Woods, Cardelús, & De Walt, 2015), which effect on two out of the three web‐building spider groups, while
would, on top of differences in connectivity, add variation in the wandering errant hunters seem to be more affected by the ver-
living conditions for epiphyte‐dwelling spiders. tical and horizontal distribution of the islands. Even though an
8 | MÉNDEZ‐CASTRO ET AL.

exhaustive analysis of these lower level patterns is outside the institutional fund to D. R. (Universidad Veracruzana, PROMEP/
scope of this paper, this issue certainly deserves further atten- 103.5/10/7748), and a Ph.D. and Postdoc scholarship to F. E. M.
tion (Mendez Castro et al., in prep) and should be taken into C. (CONACYT 307402/251146 &/251146). We are grateful to Dr.
account in future studies. Fernando Alvarez Padilla from Universidad Autonoma de Mexico
Epiphyte islands are relatively small and relatively close together (UNAM) for the advice provided in the taxonomic identification of
for most spider groups, compared to oceanic islands for most organ- the spiders.
ism groups. Still, in both these systems, it is apparent that the basic
predictors of the ETIB (island size and isolation) are scale dependent ORCID
and taxon‐specific. For example, in the Aeolian islands, cursorial
invertebrates were unable to colonize neighbouring islands, while Francisco Emmanuel Méndez-Castro http://orcid.org/0000-0002-

birds were less sensitive to isolation (Fattorini, 2010). Such specific 8926-2490

effects of isolation on organisms with different dispersal abilities


vary across geographic scales and taxonomic groups, with examples REFERENCES

coming from different insular (Cardoso et al., 2009; Carvalho & Adams, B. J., Schnitzer, S. A., & Yanoviak, S. P. (2016). Trees as islands:
Cardoso, 2014; Fattorini, 2009; Meyer & Kalko, 2008), fragmented Canopy ant species richness increases with the size of liana‐free
(Moore, Robinson, Lovette, & Robinson, 2008; Öckinger et al., trees in Neotropical forest. Ecography, 39, 1–9.
Affeld, K., Sullivan, J., Worner, S. P., & Didham, R. K. (2008). Can spatial
2009; Prugh et al., 2008), and microhabitat (Cobbold & MacMahon,
variation in epiphyte diversity and community structure be predicted
2012; Marini, Bommarco, Fontana, & Battisti, 2010) island systems. from sampling vascular epiphytes alone? Journal of Biogeography, 35,
The results of this study therefore not only support the idea of 2274–2288. https://doi.org/10.1111/j.1365-2699.2008.01949.x
epiphytic plants behaving like canopy islands but also provide Aikens, K. R., & Buddle, C. M. (2012). Small‐scale heterogeneity in tem-
perate forest canopy arthropods: Stratification of spider and beetle
another example of how the relativity of isolation (experienced dif-
assemblages. Canadian Entomologist, 144, 526–537. https://doi.org/
ferently depending on the mobility and behavioural traits of the 10.4039/tce.2012.51
target organisms) is important across taxonomic groups and spatial Angelini, C., & Silliman, B. R. (2014). Secondary foundation species as
scales. drivers of trophic and functional diversity: Evidence from tree‐epi-
phyte system. Ecology, 95, 185–196. https://doi.org/10.1890/13-
Three‐dimensional surveys of epiphytes are difficult to carry
0496.1
out, but the information about the spatial organization of epi-
Bader, M., van Dunné, H. J. F., & Stuiver, J. (2000). Epiphyte distribution
phytes provides valuable information needed for understanding in a secondary cloud forest vegetation; a case study of the applica-
their ecology and for the design of sustainable harvesting methods tion of GIS in epiphyte ecology. Ecotropica, 6, 181–195.
(Hietz, 2005; Toledo‐Aceves, Mehltreter, García‐Franco, Hernán- Barton, P. S., Evans, M. J., Foster, C. N., Cunningham, S. A., & Manning,
A. D. (2017). Environmental and spatial drivers of spider diversity at
dez‐Rojas, & Sosa, 2013). In this study, we analysed island size and
contrasting microhabitats. Journal of Ecology in the Southern Hemi-
island isolation effects on species richness, although processes sphere, 42, 700–710.
leading to this pattern, immigration, and extinction could not be Borges, P. A. V., & Hortal, J. (2009). Time, area and isolation: Factors
addressed with the single sampling used. However, due to the driving the diversification of Azorean arthropods. Journal of Biogeog-
raphy, 36, 178–191. https://doi.org/10.1111/j.1365-2699.2008.
small spatial scale and relatively fast dynamics, epiphyte‐dwelling
01980.x
biota present a very suitable system to study colonization and Cardoso, P., Aranda, S. C., Lobo, J. M., Dinis, F., Gaspar, C., & Borges, P.
extinction processes, which we recommend as a goal in further A. (2009). A spatial scale assessment of habitat effects on arthropod
research. communities of an oceanic island. Acta Oecologica, 35, 590–597.
https://doi.org/10.1016/j.actao.2009.05.005
In summary, species richness on epiphyte islands increases with
Cardoso, P., Pekár, S., Jocqué, R., & Coddington, J. A. (2011). Global pat-
island size conform the prediction of the equilibrium theory of island terns of guild composition and functional diversity of spiders. PLoS
biogeography, while isolation plays a minor role for species richness ONE, 6, e21710.
but an important role in community composition. The relativity of Carvalho, J. C., & Cardoso, P. (2014). Drivers of beta diversity in Mac-
aronesian spiders in relation to dispersal ability. Journal of Biogeogra-
isolation due to different spatial behaviour of different spider taxa as
phy, 41, 1859–1870. https://doi.org/10.1111/jbi.12348
well as the processes leading to the observed patterns should be in Chao, A., Gotelli, N. J., Hsieh, T. C., Sander, E. L., Ma, K. H., Colwell, R.
the focus of future studies. K., & Ellison, A. M. (2014). Rarefaction and extrapolation with Hill
numbers: A framework for sampling and estimation in species diver-
sity studies. Ecological Monographs, 84, 45–67. https://doi.org/10.
ACKNOWLEDGEMENTS 1890/13-0133.1
Cobbold, S. M., & MacMahon, J. A. (2012). Guild mobility affects spider
We thank members of INBIOTECA for discussions, revisions, and
diversity: Links between foraging behavior and sensitivity to adjacent
support, Verónica Hernández Arellano and Evaristo Calihua Chipa- vegetation structure. Basic and Applied Ecology, 13, 597–605.
hua for their valuable help during fieldwork, and the anonymous https://doi.org/10.1016/j.baae.2012.08.014
journal reviewers for helpful suggestions. We are grateful to the Cox, B., Moore, P. D., & Ladle, R. J. (2016). Biogeography: An ecological
and evolutionary approach (9th ed.). Hoboken, NJ: John Wiley & Sons.
brothers Martinez‐Gama (La Orduña) for permission to work on
Cruz-Angón, A., Baena, M., & Russell, G. (2009). The contribution of epi-
their coffee plantations. This work was supported by an phytes to the abundance and species richness of canopy insects in a
MÉNDEZ‐CASTRO ET AL. | 9

Mexican coffee plantation. Journal of Tropical Ecology, 25, 453–463. Méndez-Castro, F. E., & Rao, D. (2014). Spider diversity in epiphytes:
https://doi.org/10.1017/S0266467409990125 Can shade coffee plantations promote the conservation of cloud for-
De Mas, E., Chust, G., Pretus, J. Ll, & Ribera, C. (2009). Spatial modelling est assemblages? Biodiversity and Conservation, 23, 2561–2577.
of spider biodiversity: Matters of scale. Biodiversity Conservation, 18, https://doi.org/10.1007/s10531-014-0739-x
1945–1962. https://doi.org/10.1007/s10531-008-9566-2 Mestre, L., Bucher, R., & Entling, M. H. (2014). Trait‐mediated effects
Díaz, I. A., Sieving, K. E., Peña-Foxon, M. E., Larraín, J., & Armesto, J. J. between predators: Ant chemical cues induce spider dispersal. Journal
(2010). Epiphyte diversity and biomass loads of canopy emergent of Zoology, 293, 119–125. https://doi.org/10.1111/jzo.12127
trees in Chilean temperate rain forests: A neglected functional com- Meyer, C. F. J., & Kalko, E. K. V. (2008). Bat assemblages on Neotropical
ponent. Forest and Ecology Management, 8, 1490–1501. https://doi. land‐bridge islands: Nested subsets and null model analyses of spe-
org/10.1016/j.foreco.2010.01.025 cies co‐occurrence patterns. Diversity and Distributions, 14, 644–654.
Fattorini, S. (2009). The influence of geographical and ecological factors https://doi.org/10.1111/j.1472-4642.2007.00462.x
on island beta diversity patterns. Journal of Biogeography, 37, 1061– Moore, R. P., Robinson, W. D., Lovette, I. J., & Robinson, T. R. (2008).
1070. https://doi.org/10.1111/j.1365-2699.2009.02252.x Experimental evidence for extreme dispersal limitation in tropical for-
Fattorini, S. (2010). The use of cumulative area curves in biological con- est birds. Ecology Letters, 11, 960–968. https://doi.org/10.1111/j.
servation: A cautionary note. Acta Oecologica, 36, 255–258. https:// 1461-0248.2008.01196.x
doi.org/10.1016/j.actao.2009.11.005 Nakanishi, A., Sungpalee, W., Sringernyuang, K., & Kanzaki, M. (2016).
Freiberg, M., & Freiberg, E. (2000). Epiphyte diversity and biomass in the Large variations in composition and spatial distribution of epiphyte
canopy of lowland and montane forests in Ecuador. Journal of Tropi- biomass on large trees in a tropical montane forest of northern Thai-
cal Ecology, 16, 673–688. https://doi.org/10.1017/ land. Plant Ecology, 217, 1157–1169. https://doi.org/10.1007/
S0266467400001644 s11258-016-0640-7
Gonçalves, A. Z., Mercier, H., Mazzafera, P., & Romero, G. Q. (2011). Spi- Öckinger, E., Franzén, M., Rundlöf, M., & Smith, H. G. (2009). Mobility‐
der‐fed bromeliads: Seasonal and interspecific variation in plant per- dependent effects on species richness in fragmented landscapes.
formance. Annals Botany, 107, 1047–1055. https://doi.org/10.1093/ Basic and Applied Ecology, 10, 573–578. https://doi.org/10.1016/
aob/mcr047 j.baae.2008.12.002
Gonçalves-Souza, T., Almeida-Nieto, M., & Romero, G. Q. (2011). Brome- Oksanen, J., Blanchet, F. G., Friendly, M., Kindt, R., Legendre, P., McGlinn,
liad architectural complexity and vertical distribution predict spider D., … Wagner, H. (2017). Vegan: Community ecology package.
abundance and richness. Austral Ecology, 36, 476–484. https://doi. Retrieved from https://CRAN.R-project.org/package=vegan.
org/10.1111/j.1442-9993.2010.02177.x Omena, P. M., & Romero, G. Q. (2008). Fine‐scale microhabitat selection
Gonçalves-Souza, T., Brescovit, A. D., Rossa-Feres, D. C., & Romero, in a bromeliad‐dwelling jumping spider (Salticidae). Biological Journal
G. Q. (2010). Bromeliads as biodiversity amplifiers and habitat of the Linnean Society, 94, 653–662. https://doi.org/10.1111/(ISSN)
segregation of spider communities in a Neotropical rainforest. 1095-8312
Journal of Arachnology, 38, 270–279. https://doi.org/10.1636/P09- Perry, D. R. (1978). A method of access into the crowns of emergent and
58.1 canopy trees. Biotropica, 10, 155–157. https://doi.org/10.2307/
Gustafson, E. J., & Parker, G. R. (1994). Using an index of habitat 2388019
patch proximity for landscape design. Landscape and Urban Plan- Pinzon, J., Spence, J. R., & Langor, D. W. (2013). Diversity, species rich-
ning, 29, 117–130. https://doi.org/10.1016/0169-2046(94) ness and abundance of spiders (Araneae) in different strata of arbo-
90022-1 real white spruce stands. Canadian Entomologist, 145, 61–76.
Hietz, P. (2005). Conservation of vascular epiphyte diversity in Mexican https://doi.org/10.4039/tce.2012.93
coffee plantations. Conservation Biology, 19, 391–399. https://doi. Prugh, L. R., Hodges, K. E., Sinclair, A. R. E., & Brashares, J. S. (2008).
org/10.1111/j.1523-1739.2005.00145.x Effect of habitat area and isolation on fragmented animal popula-
Hsieh, T. C., Ma, K. H., & Chao, A. (2016). iNEXT: An R package for inter- tions. Proceedings of the National Academy of Sciences, 52, 20770–
polation and extrapolation of species diversity (Hill numbers). Meth- 20775. https://doi.org/10.1073/pnas.0806080105
ods in Ecology and Evolution, 7, 1451–1456. https://doi.org/10.1111/ R Development Core Team (2015). R: A language and environment for sta-
2041-210X.12613 tistical computing. Vienna, Austria: R Foundation for Statistical Com-
Jiménez-Valverde, A., Baselga, A., Melic, A., & Txasko, N. (2009). Climate puting. ISBN 3-900051-07-0. Retrieved from http://www.R-project.
and regional beta‐diversity gradients in spiders: Dispersal capacity org.
has nothing to say? Insect Conservation and Diversity, 3, 51–60. Rao, D. (2017). Habitat selection and dispersal. In C. Viera, & M. Gon-
Jocqué, R., & Dippenaar-Schoeman, A. (2006). Spiders families of the zaga (Eds.), Behaviour and ecology of spiders. Cham, Switzerland:
World. Tervuren, Belgium: Royal Museum for Central Africa. Springer.
Kindt, R., & Coe, R. (2005). Tree diversity analysis: A manual and software Revelle, W. (2017). Psych: Procedures for personality and psychological
for common statistical methods for ecological and biodiversity studies. research. Evanston, IL: Northwestern University. Retrieved from
Nairobi, Kenya: World Agroforestry Centre (ICRAF). https://CRAN.R-project.org/package=psych Version = 1.7.8.
Leroy, B., Petillon, J., Gallon, R., Canard, A., & Ysnel, F. (2012). Improving Richardson, B. A. (1999). The bromeliads microcosm and the assessment
occurrence‐based rarity metrics in conservation studies including mul- of faunal diversity in a Neotropical forest. Biotropica, 31, 321–336.
tiple rarity cut‐off points. Insect Conservation and Diversity, 5, 159– https://doi.org/10.1111/j.1744-7429.1999.tb00144.x
168. https://doi.org/10.1111/j.1752-4598.2011.00148.x Rodgers, D. J., & Kitching, R. L. (2011). Rain forest Collembola (Hexa-
MacArthur, R. H., & Wilson, E. O. (1967). The theory of island biogeogra- poda: Collembola) and the insularity of epiphyte microhabitats. Insect
phy. Princeton, NJ: Princeton University Press. Conservation and Diversity, 4, 99–106. https://doi.org/10.1111/j.
Manson, R., Hernández-Ortiz, V., Gallina, S., & Mehltreter, K. (2008). 1752-4598.2010.00104.x
Agrosistemas cafetaleros de Veracruz, biodiversidad manejo y conser- Romero, G. Q., Mazzafera, P., Vasconcellos-Neto, J., & Trivelin, P. C. O.
vación. Veracruz, Mexico: Instituto de Ecología. (2006). Bromeliad‐living spiders improve host plant nutrition and
Marini, L., Bommarco, R., Fontana, P., & Battisti, A. (2010). Disentangling growth. Ecology, 87, 803–808. https://doi.org/10.1890/0012-9658
effects of habitat diversity and area on orthopteran species with con- (2006)87[803:BSIHPN]2.0.CO;2
trasting mobility. Biological Conservation, 143, 2164–2171. https:// Scheffers, B. R., Edwards, D. P., Diesmos, A., Williams, S. E., & Evans, T.
doi.org/10.1016/j.biocon.2010.05.029 A. (2014). Microhabitats reduce animal's exposure to climate
10 | MÉNDEZ‐CASTRO ET AL.

extremes. Global Change Biology, 20, 495–503. https://doi.org/10.


1111/gcb.12439
BIOSKETCH
Southwood, T. R. E., & Kennedy, C. E. J. (1983). Trees as Islands. Oikos,
41, 359–371. https://doi.org/10.2307/3544094
Francisco E. Méndez‐Castro focusses his research on the ecol-
Stuntz, S., Ziegler, C., Simon, U., & Zotz, G. (2002). Diversity and struc-
ogy and biogeography of plant‐dwelling spiders (especially in the
ture of the arthropod fauna within canopy epiphyte species in central
Panama. Journal of Tropical Ecology, 18, 161–176. https://doi.org/10. canopy). The data presented here were obtained as part of his
1017/S0266467402002110 Ph.D. work at INBIOTECA (Universidad Veracruzana) but was
Toledo-Aceves, T., Mehltreter, K., García-Franco, J. G., Hernández-Rojas, analysed at the faculty of Geography, Marburg University.
A., & Sosa, V. J. (2013). Benefits and costs of epiphyte management
Maaike Y. Bader has a long‐standing interest in spatial ecology
in shade coffee plantations. Agriculture Ecosystems & Environment,
181, 149–156. https://doi.org/10.1016/j.agee.2013.09.026 and epiphyte ecology and more generally in understanding
Ubick, D., Paquin, P., Cushing, P. E., & Roth, V. (2005). Spiders of responses of plants in marginal habitats to environmental varia-
North America: An identification manual. American Arachnological tion and change. Glenda Mendieta‐Leiva is a postdoctoral
Society.
researcher particularly interested in ecological and biogeographi-
Wang, X., Long, W., Schamp, B. S., Yang, X., Kang, Y., Xie, Z., Xiong, M.
(2016). Vascular epiphyte diversity differs with host crown zone and cal aspects of epiphyte biology and is broadly interested in distri-
diameter, but not orientation in a tropical cloud forest. PLoS ONE, 11, butional patterns of neotropical plants. Dinesh Rao works on the
e0158548. https://doi.org/10.1371/journal.pone.0158548 behavioural ecology of spiders and is interested in the spatial
Wardle, D. A., Yates, G. W., Barker, G. M., Bellingham, P. J., Bonner, K. I.,
distribution of spider communities.
& Williamson, W. M. (2003). Island biology and ecosystem function-
ing in epiphytic soil communities. Science, 301, 1717–1720. https://d
Author contributions: F.E.M.C. and D.R. designed the data collec-
oi.org/10.1126/science.1087809
Whittaker, R. J., & Fernández-Palacios, J. M. (2008). Island biogeography: tion; F.E.M.C. and M.Y.B. conceived the ideas for the spatial data
Ecology, evolution, and conservation (2nd ed.). New York, NY: Oxford analysis; F.E.M.C. conducted the field work; F.E.M.C. and G.M.L.
University Press. did the data analysis; F.E.M.C. and M.Y.B. wrote the manuscript;
Wolff, J. O., Schneider, J. M., & Gorb, S. N. (2014). How to pass the gap
G.M.L. and D.R. revised the manuscript.
—Functional morphology and biomechanics of spider bridging
threads. In T. Asakura, & T. Miller (Eds.), Biotechnology of silk. Dor-
drecht, Netherlands: Springer.
Woods, C. L., Cardelús, C. L., & De Walt, S. J. (2015). Microhabitat
associations of vascular epiphytes in a wet tropical forest canopy. SUPPORTING INFORMATION
Journal of Ecology, 103, 421–430. https://doi.org/10.1111/1365-
2745.12357 Additional supporting information may be found online in the
World Spider Catalog (2018). Retrieved from https://wsc.nmbe.ch/. Supporting Information section at the end of the article.
Wu, L., Si, X., Didham, R. K., Ge, D., & Ding, P. (2017). Dispersal modality
determines the relative partitioning of beta diversity in spider assem-
blages on subtropical land‐bridge islands. Journal of Biogeography, 44,
2121–2131. https://doi.org/10.1111/jbi.13007 How to cite this article: Méndez-Castro FE, Bader MY,
Yanoviak, S. P., Nadkarni, N. M., & Gering, J. C. (2003). Arthropods in Mendieta-Leiva G, Rao D. Islands in the trees: A
epiphytes: A diversity component that is nor effectively sampled by
biogeographic exploration of epiphyte‐dwelling spiders.
canopy fogging. Biodiversity Conservation, 12, 731–741. https://doi.
org/10.1023/A:1022472912747
J Biogeogr. 2018;00:1–10. https://doi.org/10.1111/jbi.13422

You might also like